We're tinkering with the QoE setup in our environment, but have a bit of an issue with how we'll set it up. We've rolled the agent out to a couple machines for testing, but I'm concerned about the draw it will have on a major PROD server. As for the network tap, we can do this, but we won't be catching everything. Consider this:
My entire 'datacenter' is in a blade enclosure, running VMWare, with (2) 10GB uplinks to my core switch, which feeds to the rest of my metro-e. From client to server, about 1/3 of my clients (figure 250) feed across the metro-e and eventually flow through the uplinks, which I could tap at the switch level. These are my 'fat clients' - easy.
Now the other 2/3 of my clients connect to a load balancer, which feeds to 6 RDS servers - this is where the greater bulk of my 'user experience' is - as most of our LOB apps are running on these boxes. Problem is, they're all virtualized in my blade chassis - which basically means the comm from the RDS to any given 'app' server is at the vswitch level.
So, I realize I could use the network tap in conjunction with a few agents rolled out to critical servers (well, at least I BELIEVE that I can mix these methods) - but I'd really like to just isolate a single tap and call it a day. So, has anyone done anything with distributed vswitch, NSX? I can visualize how I would do this, just would like to know if it's been done in production and how well it works.